In [1]:
# import all libraries needed
import osmnx as ox, geopandas as gpd, pandas as pd, networkx as nx, pandas as pd
from IPython.display import Image
%matplotlib inline
ox.config(log_console=True, use_cache=True)
import scipy as sp
In [2]:
# Import distinctiveness package
from distinctiveness.dc import distinctiveness
Step 1 - Open Dataset Registerd Companies San Francisco¶
Aerial Image San Francisco Map¶
In [3]:
# open csv downloaded from San Francisco Open Data
df_buss = pd.read_csv('registered_business_sanfra.csv', low_memory=False)
df_buss.head()
Out[3]:
In [4]:
# extract specific columns needed
df_buss_2 = df_buss[['Location Id', 'DBA Name','Street Address', 'City', 'NAICS Code Description','Business Location']]
df_buss_2.head()
Out[4]:
In [22]:
# drop all rows with nan values in the columns 'NAICS Code Description' and 'Business Location'
df_buss_3 = df_buss_2[df_buss_2['NAICS Code Description'].notna()]
df_buss_4 = df_buss_3[df_buss_3['Business Location'].notna()]
df_buss_4
Out[22]:
Visualizing registered businesses in San Francisco (92693)¶
a) Classify all different categories of businesses¶
In [17]:
# undestrand the number of different categories in the dataset
# count values in the dataframe
df_buss_4['NAICS Code Description'].value_counts()
Out[17]:
In [23]:
# this for loop iterate over the dataframe and make a classification according to the type of the business category
category = []
for value in df_buss_4["NAICS Code Description"]:
if value == 'Real Estate and Rental and Leasing Services':
category.append(1)
elif value == 'Professional, Scientific, and Technical Services':
category.append(2)
elif value == 'Construction':
category.append(3)
elif value == 'Retail Trade':
category.append(4)
elif value == 'Food Services':
category.append(5)
elif value == 'Private Education and Health Services':
category.append(6)
elif value == 'Arts, Entertainment, and Recreation':
category.append(7)
elif value == 'Transportation and Warehousing':
category.append(8)
elif value == 'Accommodations':
category.append(9)
elif value == 'Financial Services':
category.append(10)
elif value == 'Certain Services':
category.append(11)
elif value == 'Wholesale Trade':
category.append(12)
elif value == 'Information':
category.append(13)
elif value == 'Administrative and Support Services':
category.append(14)
elif value == 'Multiple':
category.append(15)
elif value == 'Manufacturing':
category.append(16)
elif value == 'Insurance':
category.append(17)
elif value == 'Utilities':
category.append(18)
else:
category.append("fail")
# add new column to the dataframe
df_buss_4["Result"] = category
# print first five lines of the dataset
df_buss_4.head(10)
Out[23]:
Visualizing businesses by category in San Francisco (92693)¶
In [16]:
# export dataset in csv
df_buss_4.to_csv('buss_sanfra_categorized.csv')
Step 2 - Export Street Network (San Francisco)¶
a) Theoretical Framework¶
Source - Zachary Neal : The Connected City (fig. 5.3 'two approaches to mapping street networks')¶
b) Create Street Network using OSMnx library (by Geoff Boeing Director USC Urban Data Lab)¶
In [19]:
# define a point at the corner of California St and Mason St in SF
location_point = (37.7842, -122.4197)
# create network from point, inside bounding box of N, S, E, W each 750m from point
G2 = ox.graph_from_point(location_point, dist=3000, dist_type='bbox', network_type='drive')
G2 = ox.project_graph(G2)
fig, ax = ox.plot_graph(G2, node_size=10, node_color='white', figsize=(20,20))
Step 3 - Calculate Nodes Centrality Measures¶
In [20]:
# save street network as ESRI shapefile to work with in GIS
# specify filepath to create a folder
ox.save_graph_shapefile(G2, filepath='./shape_exported/shape_4')
G2
Out[20]:
a) Calculation Distinctiveness Centrality¶
Which node has more distinctive connections ?¶
$D_{1}(i)$ = $\sum_{j=1}^{n} w_{ij} log_{10} \frac{n - 1}{g_{j}^{a}}$
In [26]:
#Calculate the 5 metrics of Distinctiveness Centrality
DC = distinctiveness(G2, normalize = False, alpha = 1, measures=["D1"])
DC = pd.DataFrame(DC).sort_index()
DC.head()
Out[26]:
b) Calculation Degree Centrality¶
In [27]:
# perform degree centrality for the network
dgc = nx.degree_centrality(G2)
# turn dictionary into dataframe
df_dgc = pd.DataFrame.from_dict(dgc, orient='index', columns=['degree_centrality']).sort_index()
In [28]:
df_dgc.head()
Out[28]:
c) Calculation Closeness Centrality¶
In [29]:
# this line calculates closness centrality
dcc = nx.closeness_centrality(G2)
# turn dictionary into dataframe
df_dcc = pd.DataFrame.from_dict(dcc, orient='index', columns=['closeness_centrality']).sort_index()
In [30]:
df_dcc.head()
Out[30]:
d) Calculation Eigenvector Centrality¶
In [31]:
dec = nx.eigenvector_centrality_numpy(G2)
# turn dictionary into dataframe
df_dec = pd.DataFrame.from_dict(dec, orient='index', columns=['eigenvector_centrality']).sort_index()
In [32]:
df_dec.head()
Out[32]:
e) Combine All Centrality Measures¶
In [33]:
final_df = pd.concat([df_dgc, df_dcc, df_dec, DC['D1_in']], axis=1, sort=False)
final_df.head()
Out[33]:
In [59]:
# save file in csv
final_df.to_csv('nodes_metrics_business_sanfra_fixed_finaltry.csv')
Step 4 - Join operation in QGIS¶
Join Operation and Final Attribute Table¶
Step 5 - Visualize Degree Centrality and Closeness Centrality¶
Visualizing Degree Centrality Distribution¶
Visualizing Closeness Centrality Distribution¶
Step 6 - Filter nodes with highest Degree and Closeness Centrality¶
In [34]:
# open dataset with centrality and nodes coordinates exported from Qgis
full_dataset = pd.read_csv('nodes_coordinates_centrality_full.csv', low_memory=False)
full_dataset.head()
Out[34]:
In [35]:
# extract specific columns needed
full_dataset = full_dataset[['osmid', 'lon','lat', 'degree_centrality', 'closeness_centrality','eigenvector_centrality', 'D1_in']]
In [36]:
# print first five lines
full_dataset.head()
Out[36]:
a) Select nodes with highest degree centrality¶
In [37]:
# filter nodes with degree centrality bigger than 0.00259
full_dataset_high_centrality = full_dataset.loc[full_dataset['degree_centrality'] >= 0.00259]
full_dataset_high_centrality.head()
Out[37]:
In [73]:
# export in csv to visualize it in Qgis
full_dataset_high_centrality.to_csv('nodes_highest_degree_centrality.csv')
b) Create round buffer (aroound 60 meters diameter) in Qgis¶
Visualizing Nodes with Highest Degree Centrality with buffer of around 60 meters¶
c) Select nodes with highest Closeness centrality¶
In [38]:
# filter nodes with degree centrality bigger than 0.0354
full_dataset_high_closeness = full_dataset.loc[full_dataset['closeness_centrality'] >= 0.0354]
full_dataset_high_closeness.head()
Out[38]:
In [39]:
# export in csv to visualize it in Qgis
full_dataset_high_closeness.to_csv('nodes_highest_closeness_centrality.csv')
d) Create round buffer (60 meters diameter) in Qgis¶
Visualizing Nodes with Highest Closness Centrality with buffer of around 60 meters¶
Step 7 - Join Businesses Spatial Attributes with Buffers Highest Centrality¶
a) Operate Join by Attributes in Qgis combining the buffers created and the businesses locations¶
Join by location¶
Attribute table that represents the result of the join by location¶
Each osmid has all the businesses 60 meters around it¶
b) Categorize new layer coloring it with the majority of business category in it¶
Using an expression to count the majority of business typologies within the buffer¶
Step 8 - Observe if centrality measures are related to business categories¶
Visualizing Majority of Businesses Categories within Buffers Highest Degree Centrality¶
Visualizing Majority of Businesses Categories within Buffers Highest Closeness Centrality¶
Step 9 - Filtering Most Frequent Business Categories to Explore for other Patterns¶
Visualizing Majority of Businesses Categories within Buffers Highest Degree Centrality (excluding category 1 and 2)¶
Visualizing Majority of Businesses Categories within Buffers Highest Closeness Centrality (excluding category 1 and 2)¶
